1. Field of the Invention
This invention generally relates to improvements in error management in a multiplex communication system and, more particularly, to the usage of an expert system comprising techniques to identify a failing unit of a communications resource, traverse a decision tree to detect functioning members of the switch, identify each field replaceable unit (FRU) which is faulty, and circumvent the problem.
2. Description Of The Prior Art
The past several years have seen significant expansion in the use of communications multiplexers in response to the demands for increased flexibility in data communications. As these systems increased in complexity and were more tightly integrated into daily operations, outages in the communications system became less tolerable, increasing the importance of the error analysis function.
The error analysis function was done by a technician scanning through a hardware error table and sorting out the important information from the large amount of unorganized data in the error table. Often it would be necessary to manually invoke additional tests to further isolate the problem. The effectiveness of this approach depended on the degree of training and experience acquired by the technician. This approach was effective as a temporary measure; however, it was too indefinite for a permanent solution.
Another approach to improve availability of the communications system employed redundant processors. The redundant processors provided a quick recovery from a fault in one of the processors by a switchover to the other processor. One of the problems with this approach entailed the brief interval between the failure and the recovery of the second processor. During this period, a data glitch occurred. Although this was only an inconvenience during a telephone conversation, during a data communication, it frequently terminated the communication transfer. This termination necessitated the reestablishment of the communications link and the retransmittal of the data. For many applications, such as banking applications, such degradation of service is intolerable as the delivery of every message without error is essential.
The use of expert systems to diagnose patients in a medical environment is quite common today. However, these systems require the user to enter large quantities of information describing the patient as opposed to acquiring the information automatically. The user is also required to be trained as a doctor to use the system. The average technician would not be able to make effective use of the system. A more in depth discussion may be had by referring to the book edited by M. J. Coombs entitled Developments in Expert Systems, published by Academic Press (1984), and in particular to the chapter entitled "Strategic Explanations For A Diagnostic Consultation System", by Diane Hasling et al., Heuristic Programming Project, Computer Science Department, Stanford University.
An example of testing systems can be found in U.S. Pat. No. 4,601,032 to Robinson, issued July 15, 1986 and U.S. Pat. No. 4,649,515, to Thompson et al., issued Mar. 10, 1987. These patents discuss techniques for simulating faults and testing digital circuitry and process control, sensor systems. The latter employs a set of rules which respond to stimulus from outboard sensors to search a knowledge database to determine how to respond to a problem and displaying information that should guide a user to correct the problem.
Additional examples of expert systems for decision making based on a knowledge data base are disclosed in U.S. Pat. No. 4,595,982 to Burt, issued Jul. 17, 1986 and U.S. Pat. No. 4,648,044, Mar. 3, 1987. The programs described in these patents interact with a user in dialogue fashion to guide them to an answer to their problems. The tool includes question generation, legal response checking, explanation of responses and the ability to debug the knowledge data base.
While these patents describe various expert systems, they lack the capability of employing expert system techniques to identify a failing units of a communications multiplexer, traversing a decision tree to detect functioning members of the switch, identifying each unit which is faulty, and circumventing the problem.